Transcript ProgME - nslab

Samrat Ganguly et al.
IEEE JSAC’06
November 2006
Yoonchan Choi
Advanced Networking Lab
Oct 9, 2007
1

Introduction

Related Work

VoIP Service in Mesh Network

Improving the VoIP Capacity

Maintaining the VoIP Call Quality

Mobility Support for VoIP Client

Conclusion
2

Wireless VOICE-OVER-INTERNET PROTOCOL (VoIP)
 Gains significant popularity due to ubiquity of WLAN
 Requires wide area wireless coverage for true mobile phone services

IEEE 802.11-based multihop wireless mesh networks
 Considered as a practical solution for wireless VoIP service
 Has the following benefits compared with wired LAN connecting WiFi
access points
▪
▪
▪
▪
Ease of deployment and expansion
Better and wider coverage
Resilience to node failure
Reduced cost of maintenance
3

IEEE 802.11-based multihop Wireless Mesh Network for VoIP
4

Goals
 Increasing the VoIP capacity of mesh network
▪ The number of supported medium quality calls
▪ decreases with the increase in hops
▪ reduces from eight calls in single hop to one call after four hops
5

Goals
 Maintaining the VoIP quality
▪ Delay and loss characteristics can vary over time along a multihop path
between a source and destination
▪ Such variations impact the quality of a voice call
 Supporting mobility to VoIP client
▪ Maintaining calls for mobile VoIP clients that handoff to different APs during
an ongoing voice call
▪ Preventing packet losses during a handoff
6

Packet aggregation schemes
 For improving the end-to-end throughput for applications on various
types of multihop networks
 Methods
▪
▪
▪
▪
The forced algorithm
The adaptive algorithm
Packet aggregation schemes over IP-based networks
Packet aggregator for multiple VoIP streams in wireless networks
 Problems
▪ Difficult to attain the delay bound for real-time application
▪ Inefficient use of the bandwidth
▪ No consideration about the characteristics of the wireless mesh network
▪ Not adjustable for the network situation
7

VoIP Performance Metric
 R-factor
▪ Defines quality of a call
▪ Should provide a value above 70 for medium quality
R = 94.2 – 0.024d – 0.11(d – 177.3)H(d – 177.3) – 11 – 40log(1+10e)
▪ d = 25 + djitter + dnetwork
▪ e = enetwork + (1 – enetwork)ejitter
▪ H(x) = 1 if x > 0; 0 otherwise
8

VoIP Service Deployment
 Using the concept of a layer-2 switch to see the entire mesh as a single
element that switches packets between its ports
 A port is a mesh node which has at least two interfaces
▪ One in ad hoc mode for the backhaul in the mesh
▪ One in infrastructure mode to connect to clients
 Client-side transparency
▪ The clients associate with an AP using a traditional association mechanism
▪ The handoff process involves both layer-2 and layer-3 procedures
9

Testbed Hardware / Software Configuration
 Consists of 14 nodes based on the Routerboard
 200 series processor boards with 256 MB of RAM, and 512 MB of
compact flash
 Each node is equipped with two 802.11b wireless interfaces (supports
up to 4 mini-pci cards) and has an open slot for a third one (PCMCIA
16 bit)
 The testbed is spread over the third floor of NEC Research
Laboratories, Princeton, NJ
10

Mesh Node
 Using the click router on each mesh node
▪ To provide label switching, routing, and aggregation

Label-Based Forwarding
 Implementing the label-based forwarding to achieve fast path switching
▪ To support the proposed path switching
 Uses TOS field of each IP packet
▪ For packets with labels other than zero, label-based forwarding is used
▪ Packets with label zero follows underlying routing protocol used in mesh
network
 Each node has an addition table
▪ (in_label, out_label, interface, gateway)
11

VoIP Packet Aggregation
 Small sizes of packets reduce the network utilization
 Merges small voice packets from different calls into larger packets to
improve channel utilization
 Reduces the overhead and increases the number of supported calls
 Two mechanism for aggregation in a mesh network scenario
▪ End-to-end scheme
▪ The aggregation is done only at the ingress node for all flows routed for a common
destination
▪ Hop-by-hop scheme
▪ Packets are aggregated and deaggregated at every hop by adding a forced delay at
every hop
12

VoIP Packet Aggregation
P—packet being queued at a node;
A—aggregation packet being prepared;
minPackets—number of packets from the same flow that have to be aggregated at the ingress;
MTU—maximum transmission unit, in voice packets;
find queue of P;
1: if size(queue) > minPackets
add all packets from flow(P);
if size(A) < MTU
find a queue with the same dest
go to 1;
else
send A directly to destination;
else
if size(A) < MTU
add pkts w same nexthop as p;
else if aggregation timer is expired
aggregate all packets from the queue of which timer is expired also;
13

Evaluation of VoIP Packet Aggregation
 Aggregation performance for random
calls on a string topology
 Traffic statistics of three aggregation
algorithms on three calls over four-hop
leaf nodes
14

Header Compression
 A complementary scheme related to aggregation
 Packet headers with redundancy may be reduced through compression
techniques as has been done with great success for cRTP or ROHC
 Zero-length Header Compression (ZHC) algorithm
▪ Eliminates the redundant header
▪ Leverages the VoIP packet aggregation mechanism
▪ Not require context synchronization between two nodes
15

Evaluation of Header Compression
16

Routing
 To perform call admission within seconds of placing a voice call
▪ Pre-computed paths are used even if the solution is suboptimal
▪ The call admission process should preferably be distributed
 Voice call routing approach consists of
▪ Route discovery
▪ Opting for using DSDV to collect popular routes which are then pinned down and
used with label-based routing
▪ Adaptive path selection
▪ Pinning down the paths using label-based forwarding
17

Evaluation of Adaptive Routing
 Delay distribution adaptive versus fixed
▪ Only 12% of the time the delay is greater than 200 ms when the path is
adaptive
▪ A fixed path provide delays greater than 200 ms 50% of the time
18

Mobile IP-Based Scheme
 Uses a technique similar to mobile IP where each station has a unique
home AP
 Difference from Mobile IP
▪ The mobile station does not need to implement any specific protocol
 Mobile Location Register (MLR)

Flat Routing-Based scheme
 Uses a full-fledged multihop routing infrastructure in the network of
APs in the DS
19

Handoff Performance Evaluation
 Mesh topology used for handoff performance evaluation
20

Handoff Performance Evaluation
 Timeline describing a typical fast one-hop handoff for flat routing and
TMIP schemes when mobile station switches association from N1 to
N2
21

Handoff Performance Evaluation
 Handoff latencies for flat routing and TMIP with background traffic of
2.5 Mb/s and client probing on a single channel
22

Packet aggregation along with header compression can
increase the number of supported VoIP calls in a multihop
network by 2–3 times

The proposed fast path switching is highly effective in
maintaining the VoIP quality

Our fast handoff scheme achieves almost negligible disruption
during calls to roaming clients

Main contribution
 Proposing and evaluating performance optimizing techniques that are
crucial for supporting VoIP over mesh network
23